agentsopinionbearishPublic research on reasoning efficiency in AI models is severely lacking and is as important as scaling laws for RLNathan Lambert08 Aug 2026https://x.com/natolambert/status/2085552232574144570